TLDR: The United Arab Emirates has launched K2 Think, a groundbreaking open-source AI reasoning model developed by the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) and G42. Despite its compact size of 32 billion parameters, K2 Think demonstrates superior performance in complex reasoning tasks, often outperforming models more than 20 times larger. This initiative underscores the UAE’s strategic commitment to advancing AI, fostering transparency, and driving cost-effective, reproducible AI development.
Abu Dhabi, UAE – The global artificial intelligence landscape witnessed a significant advancement today with the official unveiling of K2 Think, a novel AI reasoning model developed by the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) and G42, a leading AI development holding company based in Abu Dhabi. This new open-source model is poised to redefine benchmarks in AI reasoning, offering exceptional performance in a remarkably compact form factor.
K2 Think, built with 32 billion parameters, has demonstrated its ability to outperform flagship models that are more than 20 times its size, including OpenAI’s GPT-3.5 (175B parameters) and DeepSeek-R1 (671B parameters, with 37B activated). This efficiency is a cornerstone of its design, making advanced AI reasoning more accessible and less resource-intensive. The model was developed using Alibaba’s open-source Qwen 2.5 model and is slated to be available on Cerebras Systems’ computing hardware, a key competitor to Nvidia in AI chips.
MBZUAI President and Professor Eric Xing highlighted the significance of this release, stating, ‘K2 Think, developed by MBZUAI’s Institute of Foundation Models, is a significant advancement for the global AI research and development community. By delivering these advances in a fully transparent framework, we are ushering in a new era of cost-effective, reproducible and accountable AI.’ This commitment to open-source principles allows global researchers full transparency into its training data and parameter weights, fostering collaborative innovation.
The development of K2 Think is underpinned by six pillars of innovation: long chain-of-thought supervised fine-tuning, reinforcement learning with verifiable rewards, agentic planning before reasoning, test-time scaling, speculative decoding, and inference-optimized hardware. These techniques enable the model to perform complex reasoning tasks with remarkable efficiency, as evidenced by its performance on benchmarks such as AIME24/25, GPQA-Diamond, and HMMT, where it performs on par with some of the best open-source models that require significantly more parameters.
This launch is a testament to the United Arab Emirates’ ambitious vision to solidify its position as a global leader in advanced technology and artificial intelligence. President His Highness Sheikh Mohamed bin Zayed Al Nahyan has endorsed the release, underscoring its importance as a national achievement. His Highness Sheikh Tahnoon bin Zayed Al Nahyan, Deputy Ruler of Abu Dhabi and Chairman of the Artificial Intelligence and Advanced Technology Council (AIATC), remarked, ‘The launch of K2 Think marks an important step in the UAE’s efforts to strengthen its position in advanced technologies and AI. It also reflects the country’s national vision for a future driven by innovation and knowledge.’
Also Read:
- Switzerland Launches Nationwide ‘AI Weeks’ Initiative to Foster Public Engagement and Responsible Innovation
- China’s Ascendancy in Large-Scale AI Deployment Highlighted Amidst Asia-Pacific’s Strategic Shift Towards ROI-Driven AI Adoption
K2 Think is not intended as a general-purpose chatbot but rather a specialized reasoning system, offering a path forward for building super-efficient AI reasoning models. Its introduction is expected to expand the reach of advanced AI technologies, particularly to regions with limited access to capital and infrastructure, by condensing the time and resources required for complex reasoning tasks. This initiative aligns with the UAE’s broader strategy to diversify its economy and enhance its global influence in the rapidly evolving AI sector.
